Most　Service　Function　Chains　(SFCs)　in　Network　Function　Virtualization　(NFV)　are　realized　on　the　software　or　offloading　to　the　network　interface　card　(NIC)　and　FPGA.　However,　the　software　introduces　significant　performance　overhead　while　the　NIC　and　FPGA　suffer　from　the　limited　processing　capability　and　the　development　complexity,　respectively.　In　response,　we　present　P4SC,　a　system　for　implementing　SFCs　on　the　P4-capable　device.　P4SC　provides　the　high-performance　and　flexible　SFC　implementation　by　combining　the　hardware　capability　and　the　P4　programmability.　It　offers　some　high-level　primitives　for　operators　to　build　the　SFC　requests　and　converts　the　requests　to　a　corresponding　P4　program.　Moreover,　P4SC　merges　several　SFCs　to　implement　them　on　the　same　target　while　observing　the　P4　grammar.　Besides,　P4SC　provides　the　runtime　management　of　SFCs　by　wrapping　the　low-level　device　APIs.　We　evaluate　the　P4SC　performance　on　various　P4　devices,　including　a　Barefoot　Tofino　switch.　Experimental　results　show　that　compared　to　a　state-of-the-art　NFV　framework,　P4SC　can　achieve　a　96.98%　SFC　processing　delay　decrease　on　the　P4-capable　device.